Conversions API

It is the practical implementation of server-side tracking for paid social, and it is genuinely what closes the measurement gap that opened with iOS tracking restrictions. Teams that run it well recover a real and material share of attributed conversions.

The framing worth holding on to: CAPI is a pipe. It delivers whatever you put in it, more reliably than the browser did. If what you put in it is “form submitted,” you have become more effective at teaching the platform to find people who submit forms — which is the correct outcome only if form submissions are the thing you want.

• Deduplicate with the pixel. Same event_id and event_name from both paths, or every conversion counts twice.
• Match keys decide everything. Hashed email, phone, click id, external id, IP and user agent. Meta’s event match quality score is mostly a report on how many of those you sent.
• A low match score is usually missing fields, not broken code. Check what you are sending before you debug the integration.
• Send the value. An event without a value trains the optimiser on volume, which is where this whole problem started.

The dashboard shows more conversions after the migration, which is exactly what you were told to expect, so the project is declared a success and nobody looks again.

But volume rising is the mechanical consequence of losing fewer events; it is not evidence that the campaign improved. The measurements that would tell you are downstream — MQL→SQL rate and cost per closed deal — and they are the ones nobody re-checks after a tracking migration.